Registration Number
0218U001385,
0117U003386
, R & D reports
Title
Data processing from the Large Hadron Collider in the NSC KIPT T2-center of grid-infrastructure of experiment of CMS (CERN) in 2017
popup.stage_title
Head
Levchuk Leonid Guennadievich, Кандидат фізико-математичних наук
Registration Date
07-02-2018
Organization
National Science Center "Kharkiv Institute of Physics and Technology
popup.description2
The specialized KIPT computing facility (CF) is an active working element of the Grid infrastructure of the CMS, which is one of the biggest experiments at the Large Hadron Collider (LHC) at CERN. The CF (or center T2_UA_KIPT) operates at Tier 2 (T2) of this infrastructure. 2017 was not an easy year for support of the T2_UA_KIPT center because of aging and substantial wear of the facility units and the absence of funding for the needed hardware upgrades during several previous years. Using the funds allocated within this project, the system computing productivity and disk space were increased for the first time during several years. The purchases included a disk server with an additional array of 4 TB disks and a computing ("worker") node with 10 high-performance core on each of the 2 CPU's. As a result, the mass storage disk space reached 723 TB (from 520 TB), and worker-node computing productivity was increased to 7.4 HEPspec06 (from 7.1 HEPspec06). Acquisition of some networking hardware (SFP(+) optical units, etc.) has enabled an upgrade of the system network infrastructure. To determine possible bottlenecks and sources of packet losses, the network infrastructure had been thoroughly examined. In order to provide an effective data exchange between major units, the system networking layout was completely rebuilt. The bandwidth of the main network pipes that connect system switches and routers was optimized and increased, with utilizing the aggregation technology for Ethernet channels based on the LACP ('Link Aggregation Control Protocol') and using optical units and direct-attach cables. The system uplink was also upgraded through replacement of two 1 Gbps fiber media converters by 2?10 Gbps optical units working with each other together (one on the T2_UA_KIPT gateway, and another - on the UARnet router). This has allowed us to expand the hardware limitation on the T2_UA_KIPT connectivity from 2 to 10 Gbps and substantially improve the communication quality (through, e.g., resolution of the packet loss problem). As follows from T2_UA_KIPT uplink monitoring, the information transfer rate after the upgrade is determined by CMS data and reaches 4 - 6 Gbps, with the degree of the outer channel usage being fairly large. In spite of the accomplished hardware upgrade, the T2_UA_KIPT center, by its computing and disk resources, is still well (by factor of ~3 - 3.5) below the requirements for a "nominal" T2 center of the CMS Grid infrastructure. Nevertheless, a high quality of its operation was provided during the year. The year 2017 was featured with further intensive LHC data transfers to the system for processing. For the last year, more than 1.8 PB of CMS data has been transferred to center T2_UA_KIPT, what exceeds the total amount of experimental information transferred to the site for all previous time of its participation in the CMS Grid infrastructure. (In total, more than 3.5 PB of CMS data has been transferred to the KIPT CF for all time.) In 2017, more than 310 TB of Monte Carlo (MC) data were generated at the T2_UA_KIPT center and transferred to other sites of the CMS Grid infrastructure. This is also more than amount of MC data generated at the KIPT CF for the whole period of its participation in the CMS Grid infrastructure. (In total, ~510 TB of MC data has been generated at T2_UA_KIPT and transferred to other CMS sites for all time.) The usage of KIPT CF computing resources for 11 months of 2017 was 33 thousand kHEPspec06-hours, what is some greater than was used for the total previous year 2016. This is 90.4% of the total usage of Ukrainian national Grid (NGI_UA) resources by all the virtual organizations associated with the LHC. On the other hand, this corresponds to a continuous usage of the 100% KIPT CF computing productivity during 0.53 year. A high level of quality for T2_UA_KIPT operation was provided during 2017. This quality level determined by the degree of system readiness for the distributed CMS data processing, for T2_UA_KIPT, turns out to be 97% for the year, what is one of the best results among the CMS T2 centers. So, the main tasks on support of the T2_UA_KIPT as a Tier-2 center of the CMS Grid infrastructure in 2017 have been successfully completed. An intensive work was also continued on the physics analysis of data collected in the CMS experiment. A search of SUSY particles was carried out based of the analysis of the complete event sample recorded in the previous year 2016. This data sample corresponds to 35.4 fb-1. As in the past, events with large missing transverse momentum ETmiss and a reconstructed opposite-sign high transverse momentum lepton pair were selected. Within the achieved experimental uncertainties, the corresponding SUSY signal was not observed. To accomplish this work, 401 CPU-days (4 times greater than in 2016) or 708 elapsed-time (wall-clock) days of the WLCG computing resources have been utilized. In total, 8165 analysis jobs were submitted to the WLCG/CMS infrastructure during 2017, with 73 % of this total amount being successfully completed. (These jobs were distributed practically over the entire WLCG/CMS Grid infrastructure.) In addition, ~1000 CPU hours (~40 CPU days) of the local KIPT CF computing resources were used for the analysis.
Product Description
popup.authors
Горбенко Євген Степанович
Кліменко Кім Олександрович
Ковальова Людмила Семенівна
Куров Олексій Олександрович
Левчук Леонід Геннадійович
Лук'яненко Сергій Тимофійович
Попов Віктор Пилипович
Приставка Олександр Сергійович
Сорока Дмитро В'ячеславович
popup.nrat_date
2020-04-02
